US government states open-source AI models will not face safety testing
Top line
The US government announced that open-source AI models will be exempt from voluntary safety testing, prompting calls from lawmakers for permanent regulatory legislation.
Summary
The US government has informed artificial intelligence developers that open-source AI models, such as Meta's Llama and Nvidia's Nemotron, will be exempt from voluntary safety testing. While closed models from companies like OpenAI, Google, and Anthropic face heightened scrutiny over advanced cybersecurity capabilities, the White House's decision has prompted reactions from lawmakers. Five Democratic senators recently urged the administration to work with Congress on permanent legislation for advanced domestic AI models to prevent users from turning to foreign alternatives.
Highlights
- The US government informed AI developers that open-source AI models will not be subjected to voluntary safety testing.
- Open-source models include Nvidia's Nemotron and Meta's Llama, which feature publicly accessible core components.
- Closed models are controlled by specific firms, with OpenAI, Google, and Anthropic representing the primary US developers.
- The White House discussed unpublished testing regulations with representatives from Meta, Anthropic, Google, Nvidia, and OpenAI.
- Recent disclosures by Anthropic and OpenAI regarding their AI tools breaching external systems have heightened congressional concerns over cyberattack risks.
- The administration previously stated in June that tests would be voluntary and designed for models possessing sophisticated hacking capabilities.
- Five Democratic senators urged President Trump to collaborate with Congress on permanent legislation for advanced domestic 'frontier' models.
- Lawmakers warned that unpredictable regulatory environments could drive users toward more accessible Chinese AI alternatives.
Related Items